NO FAKES REVISED, MOSTLY IMPROVED, AND PUT ON SENATE CALENDAR

By Jennifer E. Rothman
October 1, 2026

In late June a significantly revised version of NO FAKES was substituted in the Senate for the version that had been introduced in both the House (H.R. 8915) and Senate (S.4591) in late May. This newly substituted version then was moved out of the Judiciary Committee and has been placed on the Senate Legislative Calendar (Order No. 446), but given the upcoming midterm elections it is not clear whether or when it will come up for discussion. Notably, another bill that addresses in part issues of name, image, and likeness protection was just passed in the Senate—it solely focuses, however, on student athletes and primarily on a variety of broader issues related to collegiate sports. (Order No. 446, S. 4668.)

The substituted version of NO FAKES has some important and positive improvements but added at least one worrisome new change (in newly added Section 2(e)(3)) that will make the bill less effective at protecting ordinary people. Ideally, should the bill proceed, this concern and others will be addressed.

Before getting into the details of the current draft, it is worth noting a few things at the outset. 

Digital Replica Rights for All, Not Just Entertainers

This bill would create a new federal digital replica right that would apply to everyone. It was originally drafted, however, to focus primarily on the interests of record labels, and then nominally recording artists, actors, and other professional performers. The film industry has also been actively engaged in negotiating this legislation. The title of the bill itself highlights this initial impetus—NO FAKES stands for the “Nurture Originals, Foster Art, and Keep Entertainment Safe Act.” When you think about concerns over a politician appearing in a deepfake (that uses a digital replica or voice clone) that deceives the public you don’t think about fostering art or keeping entertainment safe, you think about destabilizing democracy, undermining our understanding of truth, and perhaps about possible defamation. Similarly, when you think about a teenager being digitally impersonated in a deepfake (that uses a digital replica or voice clone or other synthetic, computer-generated image or sound), the concerns are not about protecting the entertainment industry.

Despite this framing, the proposed law would apply to everyone, including politicians, teenagers and ordinary people. This is a good thing, but because of where the legislation began it doesn’t always focus on what matters— which is protecting not only recording artists, performers, actors, and athletes, but also everyone else, including ordinary people. The new draft worsens this misalignment with its added section titled “Limitations on Liability for Purely Coincidental Resemblance.” (§ 2(e)(3)—on page 88). This exception to liability provides significantly less protection to people who have not achieved “public recognition” or who do not have “commercial value.” I will discuss this in more detail in the “Key Aspects of the Bill” section.

Doesn’t Focus on Deception of the Public

The bill does not adequately protect the public from deceptive uses of digital replicas and in fact allows for deceptive uses of digital replicas so long as they are “authorized.” (For more on this concern see Reframing Deepfakes .) Some of the exceptions to liability turn on whether the uses are deceptive but most do not. Future amendments should give greater consideration to this concern.

Growing the “Identity Thicket”

The use of unauthorized digital replicas and voice clones already violates a host of state and federal laws. State laws in virtually every state protect against unauthorized uses of a person’s name, likeness, or voice for a defendant’s advantage and a host of other state laws and federal laws apply to extend liability for unauthorized uses of a person’s digital replica, including copyright, trademark, unfair competition, false advertising, false endorsement, defamation, false light, publicity, privacy, infliction of emotional distress, and fraud claims, among other laws, including more recent ones specifically targeting AI in a variety of contexts, including intimate image circulation. For more on this see Reframing Deepfakes and Grammarly Lawsuit Shows Existing Laws Can Combat Deepfakes .

Even though NO FAKES would create a new federal digital replica right, it does not preempt most existing state and federal laws that apply to and restrict unauthorized digital replicas, nor harmonize any of the existing laws. It would therefore add another layer to what I have called the “identity thicket,” in which multiple laws and entitlements conflict over who has control over a person’s name, voice, and likeness, including over digital replicas.

Platform Liability, Technology v. Law & § 230

Given that we already have a host of laws that apply here the key to meaningful change may not be more laws but improved enforcement, incentives to get platforms to remove infringing material, and technological and design changes. Technology and private cooperation can provide faster remedies than the often slow-moving legal system. Platforms can help pick up AI-generated replicas and mark them or remove them, AI models can limit the generation of unauthorized replicas of real people (as some have already done), and we can also rely on watermarks and other indicia of authenticity to certify legitimate recordings. Admittedly though, we may find ourselves in somewhat of a technological arms race with these efforts.

The shift on § 230 immunity and the notice and takedown process provided in the proposed NO FAKES may be the most impactful feature of the bill if it passes. It does, however, raise speech concerns as legitimate material may be taken down and not put back or the material may be taken down at a pivotal moment even if it is later restored. (More on that concern below.)

A simpler way to address the problem of getting platforms to take down unauthorized images and recordings (whether digital replicas or not) is to clarify that Section 230  immunity does not apply to claims arising out of the unauthorized use of a person’s name, likeness, image, voice, or other indicia of identity.  This is something about which there is a federal circuit split. 

Key Aspects of Bill

I highlight here some of the changes to the bill since its last iteration and provide a broad summary of provisions so that readers don’t need to wade back through all my prior posts on this lengthy bill (now 97 pages, although half is struck out, so substantively about 50 pages).

Creates Federal Digital Replica Right

As noted at the outset of this post, the bill would create a dedicated federal digital replica right, and would extend protection to replicas of the deceased. The bill specifically defines the right as a “property right” that is considered “intellectual property” for purposes of the Communications Act of 1934 § 230(e)(2) . This means that § 230 immunity would not apply in the context of digital replicas.

NO FAKES defines a digital replica as a “newly created, computer-generated, highly realistic electronic representation that is readily identifiable as the voice or visual likeness of an individual that—(i) is embodied in a sound recording, image, [or] audiovisual work...in which the actual individual did not actually perform or appear; or [] that is a version of a sound recording, image, or audiovisual work in which the actual individual did appear, in which the fundamental character of the performance or appearance has been materially altered[.]”

The bill excludes from the definition of a digital replica uses of samples “of one sound recording or audiovisual work into another, remixing, mastering, or digital remastering of a sound recording or audiovisual work authorized by the copyright holder.” This suggests latitude for copyright holders to engage in these standard activities without running afoul of the act.

Ownership and Control

Section 2(b) of the bill provides a “digital replication right” that gives “each individual or right holder” the right to authorize uses of digital replicas, and the voice or visual likenesses of the person in conjunction with services that produce digital replicas. Rights holders include not only the individual depicted, but also “any other individual or entity that has acquired through a license, inheritance, or otherwise, the right to authorize the use of the voice or visual likeness” of the person.  § 2(a)(6).

The digital replica right is not assignable during a person’s life. Revisions to the bill specifically note that this bar on assignments applies even in the context of bankruptcy, court orders, or similar involuntary contexts where creditors seek ownership of these digital replica rights. § 2(b)(2)(A)(i)(II). The bill, however, still allows for worrisomely long-term and broad licensing of a living person’s digital replica. This undermines protections for those depicted, as well as for the public from deception—more on this concern in a moment.

Thankfully, the most recent draft finally now requires a written authorization by the living person whose digital replica is to be created or used. Previous drafts of the legislation had allowed “authorized representatives” to enter licenses for the uses of digital replicas potentially without the actual person depicted in the digital replica knowing or approving of the license or use. Licenses involving minors (not covered by collective bargaining agreements) can be entered into by parents or guardians, but require court approval, are shorter in duration, and do not continue after a child turns 18 (though as I discuss below the replicas may be able to be used thereafter.)

There is an exception to the written-approval requirement and protections for minors, as well as the other licensing guardrails if a collective bargaining agreement (CBA) governs the use of digital replicas. § 2(b)(2)(B)(iii). Collective bargaining can help some people get better protections, as actors have gotten from SAG-AFTRA’s negotiations with film producers, but it is also possible in the future that SAG-AFTRA and other unions will not do as well in their negotiations. Additionally, most people are not represented by unions so having strong statutory protections as a floor at the outset is important both for those in unions and those who are not.

There are some other limits on the scope of licenses of digital replicas but these should be clarified and strengthened in future amendments. Specifically, licensing terms for a digital replica can be as long as ten years and digital replicas created during this time can continue to be used after the ten-year period, perhaps including in the context of new derivative works, which makes this durational limitation less meaningful than it initially appears. See § 2(b)(2)(E). And ten years is far too long a term to being with.

There are also a host of uncertain questions raised by the intersection of this law with copyright law, particularly if digital replicas are determined to be copyrightable. For example, do subsequent digital replicas infringe copyrights in earlier replicas? Does the right of reproduction and to produce derivative works make the ten-year license limit irrelevant? (My essay Copyrighting People addresses some of these questions in more depth.) 

Given these issues, meaningful limits on the scope of these licenses are crucial. The current draft requires that licenses set forth a “reasonably specific description” of the uses to be made of the digital replicas, but it is not clear how this requirement will be interpreted and it would be hard to have specific knowledge of all uses over a ten-year period (and even more so for reuses after that period). As I have previously noted, if a broad description like “use of [a specific person’s] digital replica in audio-visual works” counts as “reasonably specific” then this limitation is useless. Even if slightly more specific terms are required, such as uses of a digital replica in “promotions for [a specific] brand of soda or apparel,” the boundaries are unacceptably broad. Under either of these licenses, a person’s replica could appear in pornographic contexts or saying and doing things that the person had no awareness they were authorizing except in the most general terms. Such an outcome will work at cross-purposes with the stated objectives of protecting individuals from being exploited by AI technology and will worsen rather than protect against the deception of the public. [See Reintroduced No Fakes Act Still Needs Revision ; Reframing Deepfakes]

Postmortem Rights

The digital replica right extends to the dead, potentially for 70 years after death and applies retroactively. The term is much shorter (ten years) if a person’s identity is not used publicly prior to the two years leading up to the expiration of the initial term (and then the two years leading up to the expiration of extensions). The initial term accrues without regard for whether a person commercially exploited their identity while they were alive. There is a detailed registration process for postmortem rights with the Register of Copyrights after the initial ten-year period.

The new draft tries to address inequity in the proposed duration of the postmortem terms by linking an extension of protection to whether there is “active and authorized public use of the voice or visual likeness of the individual in interstate or foreign commerce during the 2-year period preceding the expiration” of the initial ten-year term or the subsequent 5-year terms.  It is not clear what will count as “active and authorized public use,” but the admirable goal was to provide a broader basis for extending the term then just commercialization as was the case in prior drafts. But the only specific example given in the proposed statute of meeting this standard is a commercial example; specifically, the “commercial availability of a sound recording or audiovisual work in which the voice or visual likeness of an individual is readily identifiable.” It is not clear what short of such a commercial use would count. Would the availability of a previously recorded lecture on YouTube or a university website count? Do old social media posts count if they are still publicly available? What about new posts by surviving heirs or other rights holders? For more on the vagueness of the term “commercial” in the context of IP see Commercial Speech, Commercial Use, and the Intellectual Property Quagmire.

Regardless of this uncertainty, the draft still leaves in place a system that gives more rights to those who want to exploit the dead than those who wish to limit the exploitation of the dead.  This problem is compounded—particularly for families of deceased celebrities—by the estate tax system. The IRS assesses the value of a deceased person’s name, image, likeness, and voice rights, including a potential federal digital replica right, at the highest and best use, which is understood as a commercialized use. To pay off such a tax bill, the deceased person’s identity may need to be commercialized to pay off the bill even if that is not what the deceased or their survivors would want. The proposed legislation also allows for the creation of enormous wealth arising from deceased celebrities that can be held by unrelated entities without providing justification for doing so. For a more detailed consideration of some of these issues, including consideration of appropriate terms, and other boundaries, such as with whom such postmortem rights should vest, see a recent article by myself and Anita Allen, Postmortem Privacy.

Liability

Section 2(c) of the bill provides for a civil action for the “public display, distribution, transmission, or communication of, or the act of otherwise making available to the public. . . a digital replica without authorization by the applicable right holder.” The legislation would provide liability for “distributing, importing, transmitting, or otherwise making available to the public a product or service that . . . is primarily designed to produce [] digital replicas of a specifically identified individual or individuals” without authorization. There is also liability for providing products or services that can be used to create such replicas if they have “only limited commercially significant purpose or use other than to produce” such a digital replica, or if they are “marketed, advertised, or otherwise promoted” in “concert” with an individual or entity who is providing a product or service primarily designed to produce such replicas.

Liability for defendants who are not directly distributing a digital replica and do not offer a product primarily designed to do so is limited to instances in which the defendant is on notice or has knowledge of the relevant activity. In some instances, liability is limited to instances of actual knowledge or willful avoidance as to the unauthorized nature of the replica.

Section 2(d) allows claims to be brought by any applicable rightsholder, parents and guardians, holders of recording contracts, or rightsholders to sound recordings. It is useful to allow rightsholders to enforce violations, but it would be better if future amendments clarify that rightsholders cannot sue identity-holders under the statute for using their own digital replicas. (Contract claims might exist against the identity-holder, but a person should not be able to violate their own digital replica right.)

Remedies

The bill provides the following remedies for violations:

• Statutory damages of $5,000 per work against individuals and $25,000 per work for online services acting in good faith. For online services that have not acted in good faith, statutory damages are higher—$5,000 per display, transmission, or making available, and up to $750,000 per work. For other entities, statutory damages are $25,000 per work. § 2(e)(5).
• Actual damages plus profits “attributable” to the use, other than those included in calculation of actual damages.
• Injunctive relief
• Punitive damages, if willful wrongdoing
• Fee shifting if plaintiff prevails, but only fee shifting to a prevailing defendant if the plaintiff brought case in bad faith.
• Online services with designated agents could avoid statutory damages if they reasonably believed the material was not a digital replica but could still be liable for actual damages.

Exclusions

There are many exclusions from liability (N.B. I have greatly simplified the lengthy details of each) (many, but not all, are contained within § 2(c)(4)):

• Services that provide the transmission of data from substantially all internet endpoints.
• Some online services that have user uploaded content if not “technologically or practically feasible” to disable the material or links. The “practically” modifier was added in this revision but it is unclear how this language would be interpreted and if interpreted generously could undermine the effectiveness of the statute. This language appears in a number of other places so could have a significant impact on the scope of the law.
• Nonprofit libraries or archives and their employees (acting in professional capacity).
• Accredited nonprofit educational institutions and their employees (acting in professional capacity) and when there is no “purpose of direct or indirect commercial advantage.”
• Any “individual or entity that is not an online service” if once they receive actual knowledge they act (as soon as “technologically or practically feasible”) to remove or disable material.
• Uses of digital replicas in “bona fide news, public affairs, or sports broadcast or account, provided that the digital replica is the subject of, or is materially relevant to” the broadcast or account. It is unclear whether this allows the use of digital replicas only when the replica itself is the central aspect of the story or instead allows the creation of new digital replicas to illustrate news stories and other accounts. The latter possibility could allow for deception of the public as to real events by news outlets. Perhaps this exception should have a similar limitation as the next one does for uses in documentaries and similar works that requires the work not to deceive as to its authenticity.
• Representations of the person “in a documentary or in a historical or biographical manner, including some degree of fictionalization.” This exception is limited if the use “creates the false impression that the work” is an “authentic” depiction of a work the “individual participated in” or is used as a synch track to an audiovisual work. The bill then has the increasingly common (if absurd) line that this exception to the exception doesn’t apply if the First Amendment protects the uses. Friendly reminder—the First Amendment applies whether legislators say it does or not.
• Uses consistent with public interest in bona fide commentary, criticism, scholarship, satire, or parody.
• Fleeting or negligible uses.
• Ads or promotions for many of the allowed uses.
• None of these exclusions apply if sexually explicit conduct is depicted.

The bill notes that disclaimers that a use is unauthorized or AI-generated do not relieve a party of liability. § 2(e)(4)

These exclusions go a long way to addressing many of the speech concerns raised in prior versions.

Unfortunately, a worrisome revision thrown in at the end of these exceptions in the new draft on p.88 makes it much harder for ordinary people to bring claims under the act, than for those who commercialize their identities. The new § 2(e)(3) suggests that if a “resemblance” to “the actual individual is purely coincidental” and the “actual individual has not achieved public recognition” or the person’s “voice or visual likeness” has “no commercial value” then there is no liability.  This problematically disfavors ordinary people and imports a term “commercial value” that has been particularly unclear and troublesome in litigation.  

Purely coincidental resemblance does not violate the digital replica right in the first place, so all this new provision does is raise the bar for ordinary people to bring claims and lower it for those who are already most likely to succeed and to have the resources to pursue claims. This provision is unnecessary and lays the groundwork for ordinary people to have much more limited recourse in many instances than others and for online services to continue to delay taking down unauthorized digital replicas. This late addition should be struck in future amendments (or at least equalized so applies to everyone).

Safe Harbors

Subsection 2(d) provides some safe harbors primarily for tech companies to avoid liability for their services and products, and particularly replicas created, uploaded, or circulated by others. Some of the provisions—depending on the party to whom they are directed—require a repeat infringers policy similar to that within the Copyright Act (as added by the Digital Millennium Copyright Act).  The latest version of the bill emphasizes that using tools to remove or disable content does not confer actual knowledge.

To benefit from the safe harbor, once notified (with a notification that meets statutory requirements), an online provider must remove or disable the work or link if (technologically and practically) feasible. Online services that mostly provide access to user-uploaded content should also remove other instances of the work that match a digital fingerprint of the unauthorized use and that were uploaded after the notice was processed.  There is a similar provision for platforms that primarily have sound recordings. It may be appropriate to require that the service remove all instances it could discover, not just those posted after the notice, but that doesn’t seem required as currently drafted.

There is a counter-notification process which maintains the safe harbor even if a provider puts back the material, but notably for those worried about the dangers the bill poses to free speech, there is no obligation for a service to put anything back and there may be strong incentives to keep material down. There are penalties for “false or deceptive notice[s],” including financial penalties ($25,000 per notice), damages, and costs and fees. This may help address some of the speech concerns with the takedown process, but unless there are more robust put-back requirements this is probably not sufficient.

Some have expressed concerns that the requirement of a designated agent and the detailed and onerous registration and review process required for the safe harbor will make it difficult for smaller entities to comply and compete, which would further concentrate market power in companies like Google/YouTube.

Preemption 

The preemption provision remains confounding and largely unchanged from prior versions. See § 2(g). The provision excludes from preemption any state statutes or common law in existence as of January 2, 2025 “regarding a digital replica . . . in an expressive work.”  What is meant by the language “regarding a digital replica” is unclear, which creates a significant challenge. (I will leave alone for the moment that there is not a clear definition of “expressive work” either.) Lots of state laws cover digital replicas even if they don’t use the magic phrase and even if they were passed long before we were talking about generative AI. For example, many state intimate image laws and longstanding publicity and privacy common law and statutes cover (or will be held to cover) digital replicas but were not passed or recognized specifically to do so.  Do these count as “regarding a digital replica” for purposes of the statute’s preemption provision?  It’s not clear. Depending on how this is interpreted it, it could significantly impact states that wish to pass new statutes, even broad ones like a statutory right of publicity. It is also not clear what happens if a longstanding statute is amended to more specifically address digital replicas but after the January 2, 2025 cut-off.  Similarly, it’s not clear what happens if a jurisdiction recognizes a common law right, for example of privacy or publicity, after January 2nd. In such instances, the analysis is usually based on a conclusion that the right has always existed but had not been recognized yet or, alternatively, that the claim is an appropriate evolution of preexisting common law. Would either, both, or neither of these common law determinations be preempted by the proposed federal law? Again, it’s unclear.

The preemption provision allows state laws “specifically regulating a digital replica” in “sexually explicit” or “election-related” contexts to remain in effect, but again questions loom. Do these laws have to specifically target digital replicas or can they more broadly protect against the dissemination of sexually explicit images or deception in elections? What does it mean to be “election-related”?  Does the state law need to use these specific words? Would this apply to regulation of digital replicas of politicians or only digital replicas in the context of an active election or election-related advertisements? The January 2025 cut-off also means states are treated inequitably. As drafted, the preemption provision potentially gives more authority to states like California and Tennessee (that rushed into the fray to pass specific digital replica laws) to regulate AI than other states that have taken more time to deliberate about how best to address technological changes affecting their states.

If the bill passes, the preemption provision alone is likely to generate lots of litigation and work for lawyers.

Takeaways

This version of the bill is significantly better than prior ones. Requiring the identity-holder to sign any license for their digital replica (unless subject to a CBA) is a major improvement over prior drafts. The bill also importantly would limit the transferability of a person's right to their own digital replica and place some guardrails on the scope and duration of licenses (even if those guardrails are insufficient and should be improved). The bill also would likely improve platforms willingness to remove unauthorized replicas. The new claw-back of the previously equal treatment of ordinary people added in 2(e)(3), however, is concerning and should be struck in future versions.  There should also be better protections against weaponizing the law to take down timely and important documentation of real world events of public importance. 

It’s anyone’s guess whether NO FAKES will actually get passed at this point—I hear different predictions from different folks on different days. If it stays alive, hopefully further improvements can be made both to afford better protection and to clarify the ambiguity of many terms and provisions, and particularly the uncertain interaction with numerous other state and federal laws.